One of the failings of DOS/command.com/cmd.exe is that the only logical operator is "not". The other big failing is that it can't add up. batchgen is an attempt to bring boolean logic to DOS. You need to be quite happy writing batch files before using batchgen, it doesn't quite write them for you. If you have experience of C life will be easier.

The big idea is that we use batchgen to do all the flow of control transformations, whilst we enclose normal DOS commands [# rem like this #]. batchgen turns the flow of control statements into legal DOS commands, and copies text enclosed in [##] verbatim. (this can lead to problems, see Bugs, below).

If we have the following as input:

if (%1==%2 && ! %3==%4 || exist %5)
    [# rem this will be copied verbatim
       @echo it's true #]

batchgen's output is:

@goto main
:main
if "%1" == "%2" goto AT0
goto ORF1
:AT0
if "%3" == "%4" goto ORF1
goto IF0
:ORF1
if exist %5 goto IF0
goto EI0
:IF0

     rem this will be copied verbatim
     @echo it's true
   
rem endif
:EI0
:batch_end

if we ask for optimization we get:

if not "%1" == "%2" goto ORF1
if not "%3" == "%4" goto IF0
:ORF1
if not exist %5 goto EI0
:IF0
     @echo it's true
:EI0

If you ask me, that's pretty slick, and it saves a lot of unecessary thinking. batchgen is a run of the mill lex/yacc parser. The lexemes are defined in batchgen.l, the grammar and it's associated semantic actions are defined in batchgen.y. The parser does not parse DOS commands.

batchgen [-o[r|b]] [filename]

In the absence of a filename, batchgen will read input from the command line, terminated with ctrl-z if you're using the keyboard.

batchgen filename

is equivalent to

batchgen < filename

is equivalent to

type filename | batchgen

and all three of these command lines will work. batchgen file1 file2 .. is equivalent to batchgen file1. There are no warnings or errors.

The syntax is pretty much C like.

Version 2 handles functions by generating code that uses the call :label syntax, see this. There are no changes in the grammar, but the generated code won't work unles you're on XP or later.

break may only be used inside a while or switch statement. return; returns control to the operating system. return primary_expression; (eg return 42;) is rewritten as exit /b primary_expression, see this. The errorlevel when the batch file terminates is thus set to, for example, 42.

case and default are only legal inside a switch statement, despite the grammar.

Blocks of dos code to be passed through are enclosed by [# and #]. This text is copied verbatim to the output batch file and should, therefore, be legal DOS commands. batchgen does not parse this text.

batch_file:
statement_list 
;

statement:
  dos_command 
| compound_statement 
| selection_statement 
| while_statement 
| labelled_statement 
| jump_statement 
| function_definition 
| function_call 
| 
;

function_definition:
FDEF identifier '(' ')' compound_statement 
;

function_call:
function_identifier '(' ')' 
;

labelled_statement:
  CASE primary_expression ':' statement 
| DEFAULT ':' statement 
;

compound_statement:
  '{' statement_list '}' 
| '{' '}' 
;

statement_list:
  statement 
| statement_list statement 
;

selection_statement:
  IF '(' dos_expression ')' statement 
| IF '(' dos_expression ')' statement ELSE statement 
| SWITCH '(' primary_expression ')' statement 
;

while_statement:
WHILE '(' dos_expression ')' statement 
;

jump_statement:
  BREAK ';' 
| RETURN ';' 
| RETURN primary_expression ';' 
;

dos_expression:
  logical_AND_expression 
| dos_expression LOG_OR logical_AND_expression 
;

logical_AND_expression:
  unary_expression 
| logical_AND_expression LOG_AND unary_expression 
;

unary_expression:
  dos_logical_expression 
| '(' dos_expression ')' 
| '!' unary_expression 
;

dos_logical_expression:
  ERRORLEVEL GE primary_expression 
| ERRORLEVEL < primary_expression 
| ERRORLEVEL EQ primary_expression 
| ERRORLEVEL NE primary_expression 
| EXIST primary_expression 
| primary_expression EQ primary_expression 
| primary_expression NE primary_expression 
;

primary_expression:
  identifier 
| primary_expression identifier 
| en_var 
| primary_expression en_var 
| param 
| primary_expression param 
| number 
| primary_expression number 
| empty 
;

function_identifier:
func_id 
;

You need:

compile.bat (in bin) will compile from the command line. Using MinGW produces a buggy program, you're better off using Visual Studio Express 2010 (free) and the batchgen.sln file (in vce2010).

The VC project doesn't run lex and yacc: bin\doLexYacc.bat calls lex and yacc (I use the GnuWin32 ones). Running yacc on batchgen.y gives the following error:

 state contains 1 shift/reduce conflict.

This is the if-else ambiguity, we accept yacc's default resolution.

There are some test files in tests, these should all generate and optimize without error.

This program is pretty much lifted from Compilers: Principles, Techniques and Tools by Aho, Sethi and Ullman, pub Addison Wesley 1986.

The grammar is pretty much lifted from C.